Intel Skylake / Kaby Lake

[LTC8K6]

http://browser.primatelabs.com/geekbench3/2344229

Similar numbers at just 1.6ghz?

Stepping 3 versus Stepping 2?

 

[frozentundra123456]

Statistically speaking, if you only have one result, you would expect that to be from an "average" sample. So the most valid comparison would be to compare the single value to the average of the second variable which has a large set of data.

If you are going to pick outlier values for comparison, it would be just as legitimate to compare the single value to the worst value of the other set as to compare it to the best.

 

[dahorns]

I think you're saying the same thing just differently.To which I say ~The thermals & base clock is relevant here because it's a notebook part & there's a slight chance that the turbo clocks might not last that long, highly unlikely but still, hence the hint to a performance plateau.

No, I don't think I am. Different laptop models will perform differently even if they are using the same processor. Thus, the scores you find for a processor can vary wildly based on the model of the laptop. It would be a very unreliable methodology to rate two processors by simply taking the highest scores for the two, irrespective of the model of the laptop.

On the other hand, when comparing models of phones (or models of laptops) to each other, you can more reasonably expect that any particular phone of a given model will perform similarly to another phone of that same model. Therefore, using single incident scores to compare different models should be more reliable (but not ideal) than when comparing processors across models.

I agree that thermal characteristics matter before we can make any final conclusions on Skylake's performance. My point about base clocks is simply that when running a short benchmark, turbo clocks are going to be more relevant in determining performance per clock.

 

[Sweepr]

My point about base clocks is simply that when running a short benchmark, turbo clocks are going to be more relevant in determining performance per clock.

Agreed, that's why 1.9GHz base Core i7 4500U doesn't get smoked by 2.4GHz base Core i7 5500U. It's all about Turbo in Geekbench.

 

[dahorns]

Agreed, that's why 1.9GHz base Core i7 4500U doesn't get smoked by 2.4GHz base Core i7 5500U. It's all about Turbo in Geekbench.

You see the same effect when looking at Core-M devices. Their GeekBench numbers are generally very similar to U series processors with the same max turbo.

 

[Dufus]

Different laptop models will perform differently even if they are using the same processor....

Right, and more than just that.

For example Haswell mobile quad i7-4700MQ in Geekbench 3 32-bit scores 15628 but with better cooling could maybe score over 17000.

http://browser.primatelabs.com/geekbench3/2453229

 

[Enigmoid]

Right, and more than just that.

For example Haswell mobile quad i7-4700MQ in Geekbench 3 32-bit scores 15628 but with better cooling could maybe score over 17000.

http://browser.primatelabs.com/geekbench3/2453229

That chip is around 4770k - 4790k level. I would say its overclocked.

 

[LTC8K6]

That chip is around 4770k - 4790k level. I would say its overclocked.

Yeah, is the reported speed of 2.44ghz the actual speed used when benching?:

http://ark.intel.com/products/75117/Intel-Core-i7-4700MQ-Processor-6M-Cache-up-to-3_40-GHz

 

[IntelUser2000]

http://browser.primatelabs.com/geekbench3/2344229

Similar numbers at just 1.6ghz?

Stepping 3 versus Stepping 2?

I think this is the key here folks. GeekBench doesn't do the best job of reporting clock speeds. There's no way there can be a 50%+ difference per clock between various Skylake testing.

 

[dahorns]

I think this is the key here folks. GeekBench doesn't do the best job of reporting clock speeds. There's no way there can be a 50%+ difference per clock between various Skylake testing.

GeekBench usually reports base clock. No reason base clock couldn't be easily changed.

 

[IntelUser2000]

GeekBench usually reports base clock. No reason base clock couldn't be easily changed.

Yea, that means comparing "IPC" by looking at GeekBench scores are really worthless.

You can tell by looking at numerous submissions that even identical setups can have a 20%+ difference. Macbook's Core M does not look half-bad in GeekBench, it just sucks in real world benchmarks.

 

[Dufus]

Yeah, is the reported speed of 2.44ghz the actual speed used when benching?

No, it was ~4.2GHz with some thermal throttling.

 

[PaulIntellini]

http://www.chipex.co.il/_Uploads/dbsAttachedFiles/Heterogeneouscomputerplatformenergymanagment.pdf

A very confusing presentation by Efi Rotem of Intel.

Research model have been
implemented in the next
generation Intel® Core™
processor code name Sky
Lake

Maybe this "EARTH" algorithm stuff is already in Sandy Bridge, and he just listed Skylake as a 14nm derivative.

 

[moonbogg]

So the 6700k is due august/September this year? I read that the "E" variants aren't coming until 2017. Not sure I can wait that long. This 3930k refuses to die. It won't die. Its not getting old despite being old. For Crying out loud.

 

[Dave2150]

For without them the world would be just full of us, the entitled whiners who want everything but contribute basically nothing.

You believe the only important occupation/goal in the world to be the development of new process nodes? Are you really that silly?

Get off your high horse for a moment and think about who treated you the last time you were in hospital. Or how the food need to survive magically appears at the store.

Jesus.

 

[Idontcare]

You believe the only important occupation/goal in the world to be the development of new process nodes? Are you really that silly?

Get off your high horse for a moment and think about who treated you the last time you were in hospital. Or how the food need to survive magically appears at the store.

Jesus.

Nice strawman. Rather telling how you chose to interpret my post.

 

[Dave2150]

Nice strawman. Rather telling how you chose to interpret my post.

I'm glad it is telling - I consider doctors, nurses and other helpcare workers to be of utmost importance. Whereas you believe they 'contribute nothing' - as they didn't help develop Intel's 10nm process technology.

 

[dahorns]

I'm glad it is telling - I consider doctors, nurses and other helpcare workers to be of utmost importance. Whereas you believe they 'contribute nothing' - as they didn't help develop Intel's 10nm process technology.

Seriously, crazy person, what are you talking about? His post was in jest.

 

[Ajay]

I'm glad it is telling - I consider doctors, nurses and other helpcare workers to be of utmost importance. Whereas you believe they 'contribute nothing' - as they didn't help develop Intel's 10nm process technology.

Pardon me, but you do realize that you have completely missed the context of IDC's post. Please think about it and apologize. This isn't P&N.

 

[PPB]

Nice strawman. Rather telling how you chose to interpret my post.

I get that progress in the semiconductor business can get incredibly difficult and now that we are near the end of the nanometer race it will even more than ever, but the other user had a point. You will probably find that when you get THAT inside a certain subject and it becomes your field of expertise, you can see the true depth behind what would be a trivial matter for the rest of us. Well, we can pretty much say this for many other specialized subjects where it happens too.

This doesnt invalidate the argument behind stating the true dificulty behind moving foward to new and smaller process nodes, but I cant say the same about the rest of your post. I'm sure there are plenty of people here that help the driving foward of the litho industry, it it just in ways so indirect, so broad instead of particular, and probably invisible to you and it just shows that what you probably meant for others it also applies to you. Just not on this very subject, but still.

 

[ehume]

A delay is still progress. What have any of us done to further the cause of creating or delivering 10nm? Nothing. We just sit on the sidelines bitching about it not being here on our desired schedule and at our desired cost profile.

I applaud the people who are actually elbows deep into making 10nm happen. Whenever it gets here. . . . .

Did y'all miss that part of his post? As one of those "helpcare workers," I thought his post was non-controversial. Certainly unobjectionable.

We have a delay in progress. Hey, that's engineering for you. Try building a skystalk. What? You can't? Hmm. Must be Intel's sitting on the tech, holding out . . .

 

[frozentundra123456]

Yea, dont want to put words into his mouth, but I think the point was that for those who are continually complaining, let them go out and design a better chip themselves. Kind of like the middle aged fat guy who sits on the couch all day and couldnt throw a football 20 yards, but then complains because his favorite quarterback throws an interception while a bunch of 300 pound men try to knock his head off.

We really need to get over the idea that we are entitled to a 20 or 30 or 50 percent performance gain with each generation like we used to see in the "good old days". Cpus are a mature technology now, and smaller and smaller gains are increasingly hard to come by. We need to stop and think how incredibly difficult it is to design a chip with a billion plus transistors, each barely larger than a few atoms, and mass produce millions of them with no defects.

 

[Ajay]

^ What frozentundra said.

This is a tech thread. Please try and focus. It's not OT or P&N. Go play there if you like getting your rage on (especially P&N).

 

[videogames101]

Yea, dont want to put words into his mouth, but I think the point was that for those who are continually complaining, let them go out and design a better chip themselves. Kind of like the middle aged fat guy who sits on the couch all day and couldnt throw a football 20 yards, but then complains because his favorite quarterback throws an interception while a bunch of 300 pound men try to knock his head off.

We really need to get over the idea that we are entitled to a 20 or 30 or 50 percent performance gain with each generation like we used to see in the "good old days". Cpus are a mature technology now, and smaller and smaller gains are increasingly hard to come by. We need to stop and think how incredibly difficult it is to design a chip with a billion plus transistors, each barely larger than a few atoms, and mass produce millions of them with no defects.

Fortunately for us, no one does everything. Everyone bakes their own layer of the figurative chip design cake (and a delicious cake it is!). So although you are correct that as a whole a modern CPU is an astounding and monumentally complex product/process, it's done by quite "ordinary" engineers/chemists/statisticians/etc. I too applaud the process guys neck deep in 10nm; so while I agree with IDC's intended sentiments, I also understand Dave2150's response because being neck deep in a problem is par for the course (and even "ordinary") in any sufficiently successful industry.

 

[RussianSensation]

A delay is still progress. What have any of us done to further the cause of creating or delivering 10nm? Nothing. We just sit on the sidelines bitching about it not being here on our desired schedule and at our desired cost profile.

There may be a bright side to the delay. With modern overclocked i5/i7s, you generally have a peace of mind that your underlying mobo+ram+CPU will last 4-5 years in games, maybe longer. That allows a PC user today to allocate more funds towards more fun toys, like studio monitors/speakers/audiophile headphones, new monitor tech with FreeSync/GSync or generally superior PPI monitors from what they used 5-10 years ago, faster SSDs, new graphics card purchases and games! In the past we had to be more mindful of our budgets since we more or less knew that the rate of obsolescence of our CPU platform was a ticking time bomb.

I honestly don't mind at all that CPU progress has slowed down since it means I no longer have to worry about buying and reselling my CPU+Mobo+Memory every 2-3 years which was like clock for for me for the last decade. Honestly for games it's hard to imagine that even the i7 6700K @ 4.8Ghz would be a significant upgrade from an i7 2600K @ 4.6Ghz. I just don't see it. With DX12, it might actually be better to go more cores (i.e., 6-core 4.4Ghz 5820K) rather than 4 slightly faster cores (Skylake). Unfortunately we won't have DX12 gaming benchmarks by the time Skylake launches.